CAI’s experiments

Library Environment

suppressMessages(library(tidyverse))
suppressMessages(library(stringr))
suppressMessages(library(ISLR))
suppressMessages(library(caret))
suppressMessages(library(doMC))
suppressMessages(library(plotly))
suppressMessages(library(stringr))
registerDoMC(cores=4)

Load and processing data ctu13 cleaned

myData_cleaned <- read.csv('/home/jguerra/datasets/ctu13.labeled.cleaned', stringsAsFactors = F, sep = '|')
myData_cleaned.bkp = myData_cleaned
myData_cleaned

#Periodicity
myData_cleaned = myData_cleaned %>% mutate(strong_p = str_count(State,'[a-i]'))
myData_cleaned = myData_cleaned %>% mutate(weak_p = str_count(State,'[A-I]'))
myData_cleaned = myData_cleaned %>% mutate(weak_np = str_count(State,'[r-z]'))
myData_cleaned = myData_cleaned %>% mutate(strong_np = str_count(State,'[R-Z]'))
#Duration
myData_cleaned = myData_cleaned %>% mutate(duration_s = str_count(State,'(a|A|r|R|1|d|D|u|U|4|g|G|x|X|7)'))
myData_cleaned = myData_cleaned %>% mutate(duration_m = str_count(State,'(b|B|s|S|2|e|E|v|V|5|h|H|y|Y|8)'))
myData_cleaned = myData_cleaned %>% mutate(duration_l = str_count(State,'(c|C|t|T|3|f|F|w|W|6|i|I|z|Z|9)'))
#Size
myData_cleaned = myData_cleaned %>% mutate(size_s = str_count(State,'[a-c]') + str_count(State,'[A-C]') + str_count(State,'[r-t]') + str_count(State,'[R-T]') + str_count(State,'[1-3]'))
myData_cleaned = myData_cleaned %>% mutate(size_m = str_count(State,'[d-f]') + str_count(State,'[D-F]') + str_count(State,'[u-w]') + str_count(State,'[U-W]') + str_count(State,'[4-6]'))
myData_cleaned = myData_cleaned %>% mutate(size_l = str_count(State,'[g-i]') + str_count(State,'[G-I]') + str_count(State,'[x-z]') + str_count(State,'[X-Z]') + str_count(State,'[7-9]'))
#Periodicity %
myData_cleaned <- myData_cleaned %>% mutate(strong_p = (strong_p / modelsize))
myData_cleaned <- myData_cleaned %>% mutate(weak_p = (weak_p / modelsize))
myData_cleaned <- myData_cleaned %>% mutate(strong_np = (strong_np / modelsize))
myData_cleaned <- myData_cleaned %>% mutate(weak_np = (weak_np / modelsize))
#Duration %
myData_cleaned <- myData_cleaned %>% mutate(duration_s = (duration_s / modelsize))
myData_cleaned <- myData_cleaned %>% mutate(duration_m = (duration_m / modelsize))
myData_cleaned <- myData_cleaned %>% mutate(duration_l = (duration_l / modelsize))
#Size %
myData_cleaned <- myData_cleaned %>% mutate(size_s = (size_s / modelsize))
myData_cleaned <- myData_cleaned %>% mutate(size_m = (size_m / modelsize))
myData_cleaned <- myData_cleaned %>% mutate(size_l = (size_l / modelsize))
#Making feature vectors
feature_vectors_cleaned = myData_cleaned[,c('strong_p','weak_p','weak_np','strong_np','duration_s','duration_m','duration_l','size_s','size_m','size_l','modelsize','label','class','port','proto')]
names(feature_vectors_cleaned) = c("sp","wp","wnp","snp","ds","dm","dl","ss","sm","sl","modelsize","class","subclass","port","proto")
feature_vectors_cleaned$class = factor(feature_vectors_cleaned$class)
feature_vectors_cleaned$subclass = factor(feature_vectors_cleaned$subclass)
feature_vectors_cleaned$proto = factor(feature_vectors_cleaned$proto)
feature_vectors_cleaned

Removing excesive Botnet and Normal class(Making the dataset more equitable)

feature_vectors_cleaned.bkp <- feature_vectors_cleaned
feature_vectors_cleaned %>% group_by(class) %>% summarise(n=n()) %>% arrange(desc(n))

feature_vectors_cleaned_aux_botnet <- feature_vectors_cleaned %>% filter(class == 'Botnet-TCP-SMTP-Attempt-SPAM')
feature_vectors_cleaned_aux_normal <- feature_vectors_cleaned %>% filter(class == 'Normal-TCP-HTTP')
feature_vectors_cleaned_aux_botnet

feature_vectors_cleaned_aux_normal

feature_vectors_cleaned_aux_rest <- feature_vectors_cleaned %>% filter(class != 'Botnet-TCP-SMTP-Attempt-SPAM') %>% filter(class != 'Normal-TCP-HTTP')
feature_vectors_cleaned_aux_rest %>% group_by(class) %>% summarise(n=n()) %>% arrange(desc(n))

aux1 <- rbind(feature_vectors_cleaned_aux_botnet[1:500,],feature_vectors_cleaned_aux_normal[1:500,])
aux1 %>% group_by(class) %>% summarise(n=n()) %>% arrange(desc(n))

aux <- rbind(feature_vectors_cleaned_aux_rest,aux1)
aux %>% group_by(class) %>% summarise(n=n()) %>% arrange(desc(n))

feature_vectors_cleaned <- aux

Create training set and testset

set.seed(212)
trainIndex <- createDataPartition(feature_vectors_cleaned$subclass, p=0.70, list=FALSE)
data_training <- feature_vectors_cleaned[ trainIndex,]
data_testing <- feature_vectors_cleaned[-trainIndex,]
#data_train = data_train %>% filter(length>5)
train <- upSample(x = data_training,  y = data_training$subclass, yname="class")
training <- train[,-c(11,16)]
testing <- data_testing[,-c(11)]
training

testing

nrow(training)

[1] 3322

nrow(feature_vectors_cleaned)

[1] 3089

Training configuration

ctrl_fast <- trainControl(method="cv", 
                     repeats=2,
                     number=10, 
                     summaryFunction=twoClassSummary,
                     verboseIter=T,
                     classProbs=TRUE,
                     allowParallel = TRUE)  

Experiment 1

Creation of cluster and k parameters analysis

library(factoextra)
library(cluster)
library(NbClust)
feature_vector_training = training[,-c(11,12,13,14)]
# K-means clustering
set.seed(321)
#km.res <- kmeans(feature_vector_training, 3, nstart = 25)
km.res <- kmeans(feature_vector_training, 7, nstart = 25)
# k-means group number of each observation
km.res$cluster

   [1] 2 1 2 1 1 5 5 4 2 5 5 2 3 3 3 2 1 2 3 2 2 1 5 3 2 1 3 3 3 3 3 6 3 4 2 2 1 5 2 1 7 3 2 1 1 2 1 1 1 1 3 1 3 1 1 1 3 2 2 2 2 2 2 6 2 2 2 2 6 6 5 6 4 4 2 2
  [77] 2 4 4 2 2 2 2 3 6 4 4 2 2 4 4 6 6 2 2 3 4 2 2 2 4 2 2 6 4 3 2 5 2 2 6 2 2 2 2 4 4 2 2 2 5 7 3 6 3 5 6 2 2 3 4 2 6 5 2 4 4 3 2 6 4 4 4 4 3 5 2 4 4 4 4 4
 [153] 4 6 6 4 6 3 4 5 6 4 6 3 4 4 6 4 4 4 6 2 4 4 3 3 4 6 6 5 4 6 3 4 5 6 5 4 3 4 4 4 4 4 6 6 6 6 6 6 3 3 5 4 4 1 2 2 4 2 4 6 5 4 4 2 7 6 6 6 3 6 6 7 3 4 5 6
 [229] 7 4 5 2 6 5 3 1 1 1 6 4 6 4 6 6 5 6 6 6 6 6 6 6 7 5 7 5 1 3 2 1 2 1 2 1 1 1 1 2 2 3 2 5 5 5 1 7 7 7 7 1 3 1 1 2 3 2 2 3 3 1 1 1 1 2 2 1 3 1 1 2 2 3 4 3
 [305] 1 1 1 2 2 3 1 2 1 3 1 2 1 2 1 2 3 2 2 2 2 2 1 1 2 5 5 1 1 5 1 1 1 1 1 1 1 1 1 1 1 6 1 1 6 2 1 1 1 5 2 1 2 5 2 2 1 4 1 6 2 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1
 [381] 1 1 1 1 1 1 1 1 1 3 2 2 1 1 3 2 1 2 2 2 2 5 1 3 5 3 2 1 2 1 2 2 1 1 1 1 3 1 2 1 5 2 2 1 1 1 3 3 1 1 2 2 2 1 3 2 2 1 1 3 1 2 6 2 6 5 5 5 5 2 2 2 2 2 2 3
 [457] 2 2 2 2 2 2 2 6 1 6 6 6 3 3 6 6 3 6 6 2 1 1 2 1 2 6 4 1 3 2 2 2 3 1 3 2 3 1 1 1 1 1 1 2 2 1 2 1 2 3 1 1 6 2 3 1 1 1 7 1 7 2 3 1 3 3 3 2 2 3 2 1 1 1 1 3
 [533] 1 2 1 2 5 7 4 3 3 3 1 1 2 2 1 2 3 3 2 2 2 1 1 3 1 5 1 3 2 1 1 1 5 1 2 2 3 2 2 2 3 3 3 3 1 2 1 2 2 2 3 3 2 3 1 3 1 1 1 2 1 1 1 1 3 3 3 2 2 1 2 2 2 2 1 1
 [609] 1 6 2 1 1 2 1 1 1 3 1 1 2 2 3 2 1 1 1 1 3 3 1 1 4 2 2 2 1 2 3 2 2 1 5 2 2 1 2 2 1 1 1 2 3 2 2 3 3 3 3 1 1 1 1 1 2 2 2 3 1 1 1 2 2 3 2 1 1 1 3 1 1 4 1 2
 [685] 3 3 2 1 3 3 1 1 1 2 3 5 1 1 1 5 2 2 2 3 1 3 3 1 2 1 1 2 2 2 3 2 1 3 3 1 7 2 3 1 3 2 7 1 2 1 1 2 1 1 1 2 2 3 1 3 1 1 2 1 1 1 3 1 3 2 1 1 1 3 2 1 3 2 4 1
 [761] 1 3 1 2 3 2 1 1 1 1 1 2 1 4 3 3 2 1 1 2 3 2 3 1 1 2 3 2 2 2 2 3 7 2 3 1 3 6 3 1 2 2 2 3 1 1 1 3 1 1 1 1 1 1 1 2 3 2 1 4 1 1 1 3 1 2 1 4 3 3 1 6 2 3 1 1
 [837] 2 1 2 1 1 1 2 1 1 1 1 1 4 1 2 1 7 1 1 1 1 1 2 2 1 1 1 2 1 1 2 2 3 1 1 2 1 2 2 1 1 2 3 3 3 2 3 3 1 1 1 1 2 1 1 2 2 2 1 1 3 1 1 1 1 2 1 3 3 2 2 2 2 1 2 1
 [913] 3 1 5 2 1 1 1 1 2 2 4 2 1 1 3 2 2 2 1 3 2 1 1 1 3 3 2 3 3 1 1 1 2 1 1 1 1 3 3 1 3 3 2 1 1 1 1 1 2 2 1 2 1 3 2 1 2 2 3 2 2 2 5 3 2 2 2 5 1 1 1 1 1 2 2 1
 [989] 1 1 1 3 1 1 1 3 1 2 1 3
 [ reached getOption("max.print") -- omitted 2322 entries ]

# Visualize k-means clusters
fviz_cluster(km.res, data = feature_vector_training, geom = "point",
             stand = FALSE, ellipse.type = "norm")

Elbow analysis

set.seed(321)
# Compute and plot wss for k = 2 to k = 15
k.max <- 15 # Maximal number of clusters
data <- feature_vector_training
wss <- sapply(1:k.max, 
        function(k){kmeans(data, k, nstart=10 )$tot.withinss})
plot(1:k.max, wss,
       type="b", pch = 19, frame = FALSE, 
       xlab="Number of clusters K",
       ylab="Total within-clusters sum of squares")
abline(v = 3, lty =2)

Silhouette analysis

set.seed(322)
k.max <- 10
data <- feature_vector_training
nrow(data)

[1] 3322

sil <- rep(0, k.max)
# Compute the average silhouette width for 
# k = 2 to k = 15
for(i in 2:k.max){
  km.res <- kmeans(data, centers = i, nstart = 25)
  ss <- silhouette(km.res$cluster, dist(data))
  sil[i] <- mean(ss[, 3])
}
# Plot the  average silhouette width
plot(1:k.max, sil, type = "b", pch = 19, 
     frame = FALSE, xlab = "Number of clusters k")
abline(v = which.max(sil), lty = 2)

Useful functions

Data training partitions: cold start study

Iteration #1

output_1 <- result$output
output_t_1 <- result$output_t
output_1

gg <- ggplot(data = output_1)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

first_training_sample <- training.sampled_1[1:200,]
ggplot(first_training_sample) + geom_bar(aes(subclass))

class_distribution <- first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

testing_result <- result$testing_result
#testing_result
gg <- ggplot(data = output_t_1)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

output_t_aux_1 <- output_t_1
names(output_t_aux_1) <- c('data_count_t','metric_t')
output_result_1 <- cbind(output_1,output_t_aux_1)
gg <- ggplot(data = output_result_1)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #2

output_2 <- result_2$output
output_t_2 <- result_2$output_t
output_2

gg <- ggplot(data = output_2)
  gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
    labs(title="Random Forest through data training size", 
         #subtitle="Drawn from Long Data format", 
         caption="Source: CTU-13", 
         y="F1 Score", 
         color=NULL)

first_training_sample <- training.sampled_2[1:200,]
ggplot(first_training_sample) + geom_bar(aes(subclass))

class_distribution <- first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

testing_result_2 <- result$testing_result
#testing_result
gg <- ggplot(data = output_t_2)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

output_t_aux_2 <- output_t_2
names(output_t_aux_2) <- c('data_count_t','metric_t')
output_result_2 <- cbind(output_2,output_t_aux_2)
gg <- ggplot(data = output_result_2)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #3

output_3 <- result_3$output
output_t_3 <- result_3$output_t
output_3

gg <- ggplot(data = output_3)
  gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
    labs(title="Random Forest through data training size", 
         #subtitle="Drawn from Long Data format", 
         caption="Source: CTU-13", 
         y="F1 Score", 
         color=NULL)

first_training_sample <- training.sampled_3[1:200,]
ggplot(first_training_sample) + geom_bar(aes(subclass))

class_distribution <- first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

testing_result_3 <- result$testing_result
#testing_result
gg <- ggplot(data = output_t_3)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

output_t_aux_3 <- output_t_3
names(output_t_aux_3) <- c('data_count_t','metric_t')
output_result_3 <- cbind(output_3,output_t_aux_3)
gg <- ggplot(data = output_result_3)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #4

output_4 <- result_4$output
output_t_4 <- result_4$output_t
output_4

gg <- ggplot(data = output_4)
  gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
    labs(title="Random Forest through data training size", 
         #subtitle="Drawn from Long Data format", 
         caption="Source: CTU-13", 
         y="F1 Score", 
         color=NULL)

first_training_sample <- training.sampled_4[1:200,]
ggplot(first_training_sample) + geom_bar(aes(subclass))

class_distribution <- first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

testing_result_4 <- result$testing_result
#testing_result
gg <- ggplot(data = output_t_4)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

output_t_aux_4 <- output_t_4
names(output_t_aux_4) <- c('data_count_t','metric_t')
output_result_4 <- cbind(output_4,output_t_aux_4)
gg <- ggplot(data = output_result_4)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #5

output_5 <- result_5$output
output_t_5 <- result_5$output_t
output_5

gg <- ggplot(data = output_5)
  gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
    labs(title="Random Forest through data training size", 
         #subtitle="Drawn from Long Data format", 
         caption="Source: CTU-13", 
         y="F1 Score", 
         color=NULL)

first_training_sample <- training.sampled_5[1:200,]
ggplot(first_training_sample) + geom_bar(aes(subclass))

class_distribution <- first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

testing_result_5 <- result$testing_result
#testing_result
gg <- ggplot(data = output_t_5)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

output_t_aux_5 <- output_t_5
names(output_t_aux_5) <- c('data_count_t','metric_t')
output_result_5 <- cbind(output_5,output_t_aux_5)
gg <- ggplot(data = output_result_5)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Data training partitions: cold start study (simple Random Forest)

Iteration #1

rf_output_1 <- rf_result_1$output
rf_output_t_1 <- rf_result_1$output_t
rf_output_1

gg <- ggplot(data = rf_output_1)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_first_training_sample <- rf_training.sampled_1[1:200,]
ggplot(rf_first_training_sample) + geom_bar(aes(subclass))

rf_class_distribution <- rf_first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(rf_first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

#rf_testing_result_1 <- rf_result_1$testing_result
#testing_result
gg <- ggplot(data = rf_output_t_1)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_output_t_aux_1 <- rf_output_t_1
names(rf_output_t_aux_1) <- c('data_count_t','metric_t')
rf_output_result_1 <- cbind(rf_output_1,rf_output_t_aux_1)
gg <- ggplot(data = rf_output_result_1)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #2

rf_output_2 <- rf_result_2$output
rf_output_t_2 <- rf_result_2$output_t
rf_output_2

gg <- ggplot(data = rf_output_2)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_first_training_sample <- rf_training.sampled_2[1:200,]
ggplot(rf_first_training_sample) + geom_bar(aes(subclass))

rf_class_distribution <- rf_first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(rf_first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

#rf_testing_result_2 <- rf_result_2$testing_result
#testing_result
gg <- ggplot(data = rf_output_t_2)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_output_t_aux_2 <- rf_output_t_2
names(rf_output_t_aux_2) <- c('data_count_t','metric_t')
rf_output_result_2 <- cbind(rf_output_2,rf_output_t_aux_2)
gg <- ggplot(data = rf_output_result_2)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #3

rf_output_3 <- rf_result_3$output
rf_output_t_3 <- rf_result_3$output_t
rf_output_3

gg <- ggplot(data = rf_output_3)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_first_training_sample <- rf_training.sampled_3[1:200,]
ggplot(rf_first_training_sample) + geom_bar(aes(subclass))

rf_class_distribution <- rf_first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(rf_first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

#rf_testing_result_2 <- rf_result_2$testing_result
#testing_result
gg <- ggplot(data = rf_output_t_3)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_output_t_aux_3 <- rf_output_t_3
names(rf_output_t_aux_3) <- c('data_count_t','metric_t')
rf_output_result_3 <- cbind(rf_output_3,rf_output_t_aux_3)
gg <- ggplot(data = rf_output_result_3)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #4

rf_output_4 <- rf_result_4$output
rf_output_t_4 <- rf_result_4$output_t
rf_output_4

gg <- ggplot(data = rf_output_4)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_first_training_sample <- rf_training.sampled_4[1:200,]
ggplot(rf_first_training_sample) + geom_bar(aes(subclass))

rf_class_distribution <- rf_first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(rf_first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

#rf_testing_result_2 <- rf_result_2$testing_result
#testing_result
gg <- ggplot(data = rf_output_t_4)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_output_t_aux_4 <- rf_output_t_4
names(rf_output_t_aux_4) <- c('data_count_t','metric_t')
rf_output_result_4 <- cbind(rf_output_4,rf_output_t_aux_4)
gg <- ggplot(data = rf_output_result_4)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Iteration #5

rf_output_5 <- rf_result_5$output
rf_output_t_5 <- rf_result_5$output_t
rf_output_5

gg <- ggplot(data = rf_output_5)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_first_training_sample <- rf_training.sampled_5[1:200,]
ggplot(rf_first_training_sample) + geom_bar(aes(subclass))

rf_class_distribution <- rf_first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(rf_first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

#rf_testing_result_2 <- rf_result_2$testing_result
#testing_result
gg <- ggplot(data = rf_output_t_5)
gg + geom_line(aes(x = data_count, y = metric),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

rf_output_t_aux_5 <- rf_output_t_5
names(rf_output_t_aux_5) <- c('data_count_t','metric_t')
rf_output_result_5 <- cbind(rf_output_5,rf_output_t_aux_5)
gg <- ggplot(data = rf_output_result_5)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + geom_line(aes(x = data_count_t, y = metric_t),color = 'blue') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="F1 Score", 
       color=NULL)

Studies Samples

first_training_sample <- training.sampled[1:200,]
first_training_sample
ggplot(first_training_sample) + geom_bar(aes(subclass))

class_distribution <- first_training_sample %>% group_by(class) %>% summarise(n = n()) %>% arrange(desc(n))
ggplot(first_training_sample) + geom_bar(aes(class)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

part 2

set.seed(206)
library(doParallel)
cl <- makeCluster(2)
registerDoParallel(cl)
size_training <- nrow(training)
split_size_training = size_training / 200
count_random <- foreach(i=1:split_size_training) %dopar% {
  200 * i
}
training.sampled <- training[sample(size_training, size_training), ]
metric <- foreach(i=1:split_size_training) %do% {
  #library(caret)
  count <- 200 * i
  aux_training_set <- training.sampled[c(1:count), ]#training[sample(size_training, count), ]
  clusters <- kmeans(aux_training_set[,-c(11,12,13,14)],3,nstart = 25)
  aux_training_set_cluster <- cbind(aux_training_set, cluster = clusters$cluster)
  result_vector <- numeric(nrow(testing))
  for (j in c(1:3)){
    cluster_data <- filter(aux_training_set_cluster, cluster == j)
    new_rfFit <- train(class ~ sp+wp+wnp+snp+ds+dm+dl+ss+sm+sl,
               data = cluster_data,
               metric="ROC",
               method = "rf",
               trControl = ctrl_fast)
    predsrfprobs <- predict(new_rfFit,testing,type='prob')
    for (k in c(1:length(result_vector))){
      if(predsrfprobs$Botnet[k] > 0.5){
        result_vector[k] <- result_vector[k] + 1
      }
      else{
        result_vector[k] <- result_vector[k] - 1
      }
    }
    
  }
  a = ifelse(result_vector > 0,'Botnet','Normal')
  cm <- confusionMatrix(a,testing$class)
  metric <- cm$byClass['F1']#cm$overall[1]
  metric
  
}

output <- do.call(rbind, Map(data.frame, data_count=count_random, metric=metric))
output
gg <- ggplot(data = output)
gg + geom_line(aes(x = data_count, y = metric),color = 'red') + 
  labs(title="Random Forest through data training size", 
       #subtitle="Drawn from Long Data format", 
       caption="Source: CTU-13", 
       y="Accuracy", 
       color=NULL)
cluster_data

Test with only one

set.seed(226)
size_training <- nrow(training)
training.sampled <- training[sample(size_training, size_training), ]

aux_training_set <- training.sampled[c(1:200), ]#training[sample(size_training, 200), ]
clusters <- kmeans(aux_training_set[,-c(11,12,13,14)],3,nstart = 25)
aux_training_set_cluster <- cbind(aux_training_set, cluster = clusters$cluster)
result_vector <- numeric(nrow(testing))
result_vector_trainning <- numeric(nrow(aux_training_set))

for (j in c(1:3)){
  cluster_data <- aux_training_set_cluster %>% filter(cluster == j)
  new_rfFit <- train(subclass ~ sp+wp+wnp+snp+ds+dm+dl+ss+sm+sl,
               data = cluster_data,
               metric="ROC",
               method = "rf",
               trControl = ctrl_fast)
  predsrfprobs <- predict(new_rfFit,testing,type='prob')
  for (k in c(1:length(result_vector))){
    if(predsrfprobs$botnet[k] > 0.5){
      result_vector[k] <- result_vector[k] + 1
    }
    else{
      result_vector[k] <- result_vector[k] - 1
    }
  }
  
  #Trainning predict
  predsrfprobs_t <- predict(new_rfFit,aux_training_set,type='prob')
  for (k in c(1:length(result_vector_trainning))){
    if(predsrfprobs_t$botnet[k] > 0.5){
      result_vector_trainning[k] <- result_vector_trainning[k] + 1
    }
    else{
      result_vector_trainning[k] <- result_vector_trainning[k] - 1
    }
  }
}

a = ifelse(result_vector > 0,'botnet','normal')
b <- ifelse(result_vector_trainning > 0,'botnet','normal')
cm <- confusionMatrix(a,testing$subclass)
metric <- cm$byClass['F1']#cm$overall[1]
metric
cm_t <- confusionMatrix(b,aux_training_set$subclass)
metric_t <- cm_t$byClass['F1']
metric_t

Sample examples

set.seed(556)
a = c(1,2,3,4,5,6,7,8,9)
r <- sample(9,3)
a[r]
r2 <- sample(9,3)
a[r2]

#testing_result
testing_result.bkp <- testing_result
testing_result
names_aux <- foreach(i=1:(nrow(training)/200)) %do% {
    iteration <- 200 * i
    paste('size_',toString(iteration),sep = "")
}
testing_result_names <- unlist(names_aux, use.names=FALSE)
testing_result <- testing_result[,c(-1)]
names(testing_result) <- testing_result_names
testing_result

testing_aux <- cbind(testing,testing_result)
testing_aux.bkp2 <- testing_aux
#write.table(testing_aux,file="testing_cluster_result.txt",sep="|", row.names = F)
testing_aux
sums <- rowSums(testing_aux[,-c(1:14)])
sums
testing_aux[,-c(1:14)]
testing_aux <- cbind(testing_aux,sums)
testing_aux
testing_aux_result <- testing_aux %>% group_by(class) %>% summarise(n = n(), sums = sum(sums)) %>% arrange(desc(sums))
testing_aux_result

graph_testing_result <- ggplot(testing_aux_result[-c(1,nrow(testing_aux_result)),])
graph_testing_result + geom_point(aes(class,sums)) + theme(axis.text.x = element_text(angle = 90, hjust = 1))

feature_vectors_cleaned

library(gridExtra)
pdf("data_output.pdf", height=11, width=8.5)
grid.table(feature_vectors_cleaned[1:20,])
dev.off()

testing_result.bkp
testing_aux.bkp2
testing_aux_result

rusty_data_result <- testing_aux.bkp2
rusty_data_result_short <- rusty_data_result[,-c(1:11,14)]
rusty_data_result_short[,-c(1,2)]
rusty_data_result_short$pos <- rowSums(rusty_data_result_short[,-c(1,2)] > 0)
rusty_data_result_short$neg <- rowSums(rusty_data_result_short[,-c(1,2)] < 0)
rusty_data_result_short_cleaned <- rusty_data_result_short[,c(1,2,46,47)]
rusty_data_result_short_cleaned
rusty_data_result_short_cleaned_result <- rusty_data_result_short_cleaned %>% mutate(good = ifelse(subclass == 'normal',neg,pos))
rusty_data_result_short_cleaned_result <- rusty_data_result_short_cleaned_result %>% mutate(bad = ifelse(subclass == 'normal',pos,neg))
rusty_data_result_short_cleaned_result %>% group_by(port) %>% summarise(n=n(),good = sum(good),bad = sum(bad)) %>% arrange(desc(n))

data_botnet_port <- rusty_data_result_short_cleaned_result %>% filter(subclass == 'botnet')
data_normal_port <- rusty_data_result_short_cleaned_result %>% filter(subclass == 'normal')
data_botnet_port_result <-  data_botnet_port %>% group_by(port) %>% summarise(n=n(),good = sum(good),bad = sum(bad)) %>% arrange(desc(n))
data_normal_port_result <- data_normal_port %>% group_by(port) %>% summarise(n=n(),good = sum(good),bad = sum(bad)) %>% arrange(desc(n))
data_botnet_port_result
data_normal_port_result

ggplot(data = data_botnet_port_result) + 
  geom_bar(mapping = aes(x = port, fill = clarity))

#write.table(data_botnet_port_result,file="data_botnet_port.txt",sep="|", row.names = F)
library(reshape2)
data <- data_botnet_port_result
data$port <- as.factor(data$port)

melt(data[,c(1,3,4)])

ggplot(melt(data[,c(1,3,4)]))+
  geom_col(aes(x=port,y=value,fill=variable))+
  #theme_bw()+
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

Making Noisy data (training_noisy: dataset to train with 20% of noisy)

set.seed(101) 
training.bkp <- training
noisy_data <- training

porcent <- nrow(training) / 5
training_noisy <- generate_data_noisy(noisy_data,porcent)
nrow(training)
nrow(training_noisy)

cm$overall[1]

 Accuracy 
0.9222042 

Robustness Analisys

set.seed(321)
partitions <- c(1:15)
rla_measure_result <- c()
total <- nrow(training)
balanced_accuracy <- randomForest_performace(training,testing,'Balanced Accuracy')

for(i in partitions){
  porcent <- (i*total) / 100
  training_noisy <- generate_data_noisy(noisy_data,porcent)
  balanced_accuracy_aux <- randomForest_performace(training_noisy,testing,'Balanced Accuracy')
  rla_measure_result[i] <- balanced_accuracy_aux
}
rla_measure_result

plot(rla_measure_result)

set.seed(322)
partitions <- seq(2,30,2)
rla_measure_result_2 <- c()
total <- nrow(training)
#balanced_accuracy <- randomForest_performace(training,testing,'Balanced Accuracy')

for(i in partitions){
  porcent <- (i*total) / 100
  training_noisy <- generate_data_noisy(noisy_data,porcent)
  balanced_accuracy_aux <- randomForest_performace(training_noisy,testing,'Balanced Accuracy')
  rla_measure_result_2[i] <- balanced_accuracy_aux
}
rla_measure_result_2

plot(rla_measure_result_2)